simtk.org

Hi there,

I have a question regarding RRA.

For context- I have collected experimental gait data (i.e., marker and GRF) on 14 subjects. Each subject completed three different conditions: level ground, incline, and decline. I have six successful trials for each condition (three for each leg). This means I have 18 trials in total to analyze for each subject.

I have scaled my models, completed IK and ID and intend to proceed to SO. I am currently working through RRA and came across a forum post discussing the option 1) running RRA on every trial (i.e., 18 trials) for every subject vs. option 2) running RRA on each condition (i.e., 3 conditions) for every subject. Here is the post:
viewtopicPhpbb.php?f=91&t=17930&p=49934&start=0&view=

This made me wonder how someone might proceed with option 2). I agree with the forum post that running RRA on every trial (option 1) contradicts the idea that our subject will not be changing between each trial. For this reason, I think it would make more sense to run RRA on one trial for each condition and then use this adjusted model for the remaining analyses of that subject (in that condition).

However, this is where my question comes in.... Outputs of RRA include the adjusted model and adjusted kinematics. Normally when proceeding to SO, I would input the adjusted model and the adjusted kinematics into my SO analysis after doing RRA. However, in the case of option 2), I would not have adjusted kinematics for every single trial. What I'm wondering is would it make sense to proceed with SO, using the one adjusted model per condition and the original kinematics (from IK) for all trials of that condition?

Thank you for your help! I appreciate it.

Tess

Tess,

RRA is needed before doing CMC, but Static Optimization (SO) actually works fine without it.

You may be doing RRA because the residual loads seem high, but the RRA could easily make things worse for you. For two reasons:

1. If the residuals are large, that is an indication of error in model and/or data. Which is important to know, and RRA might hide such issues. With walking on slopes, the force plate data needs a coordinate transformation, and gravity compensation, before it can be entered into the inverse dynamic (ID) analysis. Large residuals could be an indication that this has not been done correctly.

2. If there are no problems with the data, residuals will still exist. Most of the body mass is in the upper body, so dynamic residuals are mostly caused by model and marker set not representing the upper body well. However, your lower extremity analysis will still be fine, because those results do not depend on upper body motion and data. RRA would make things worse if it partially (incorrectly) blames the leg model and leg motion for the residuals, and starts making changes there. Lower extremity motion capture and lower extremity models are quite reliable and RRA would generally not result in improvements. Don't take my word for it, though. To prove that ID/SO results don't depend on the upper body, make a large change to the upper body mass and observe that the results are not affected.

My suggestion would be to proceed without RRA, but first make sure that the data is valid.

Ton van den Bogert